Hacker News new | past | comments | ask | show | jobs | submit login
[flagged] Stereotype accuracy: One of the largest, most replicable effects in psychology [pdf] (gwern.net)
67 points by feross 38 days ago | hide | past | favorite | 57 comments

We are walking neural networks. Pattern recognition is what we do best. Stereotypes are recognized patterns.

Are they 100% accurate? No way. But they provide good starting point for making decisions, especially quick decisions your life depends on.

E.g. should I continue walking on this side of the road when there is a homeless (recognized by cloths) guy ahead holding a knife?

It is possible that he is a nice person and maybe not even homeless, just having very casual style? Holding knife that he just found and trying to find it’s owner as a good citizen?


But for slim chance that he is a homeless schitzophrenic that going to slash illusionary monsters approaching him with that knife, I would cross the road and walk on another side.

Worse, we need to keep in mind that stereotypes can be totally wrong if they aren't inferring the correct causality. If we have three observables P, Q, R, and physical causations P -> Q and P -> R, then stereotype-style pattern inference will incorrectly conclude either Q -> R or R -> Q.

More generally, stereotype-oriented reasoning has no place for the argument from the outside that, when we mechanically disprove a stereotype's existence, we must refute the entire chain of correlations which led to the stereotype's establishment.

And of course worth noting, relying on them is a bad shortcut when there is time and resources to gather more data.

Pretty specific stereotype you got there with the homeless schitzophrenic.

I don't think anyone is worried about stereotypes being applied when determining how far to get away from someone with a knife, though I imagine black people experience more roads crossings than normal, knife or not.

"The problem with stereotypes is not that they are untrue, but that they are incomplete. They make one story become the only story.

"The consequence of the single story is this: It robs people of dignity. It makes our recognition of our equal humanity difficult. It emphasizes how we are different, rather than how we are similar."

     -- Author Chimamanda Ngozi Adiche

If stereotypes are accurate does it follow that discriminating on their basis is legitimate? Or can they be accurate patterns that we are obliged to ignore? Must we make no assumptions and treat every stranger as a blank slate to be ethical? Or is it sufficient to update strongly on specific observations as they come?

If stereotypes are accurate then the optimal amount of prejudice is greater than zero, as long as it is outweighed by postjudice.

I grew up being taught that sterotypes are a myth. Everyone is the same.

Boy I was in for a rude awakening

Regardless of how accurate it may or may not be, any blanked generalisation over millions of people is in fact useless for individual judgements. It is quite dehumanising as well, seeing it's people we're talking about.

I've had people assume traits about me due to my place of birth. I did not like it one bit I can assure you, hence I strive not to do so with others.

It's also worth noting that the article found that national stereotypes are the exception from the rule and a very mixed bag (for examples Russians having somewhat correlating stereotypes of Finns, but Finns having completely false stereotypes of Russians). Political, racial, or gender stereotypes are pretty reliable (according to the article), but national stereotypes are better ignored.

What a radical document. I doubt the comment thread will turn into anything more than a flaming mess; but this does go to one of the more interesting parts of the scientific process.

Scientists are, by and large, going to be well off. It is very hard to spend a life studying stuff that nobody understands while also been dirt poor or otherwise on the margins of society. Possible, perhaps, but not feasible on average.

Scientists therefore tend to be a little more well off than the average, a little smarter and a lot more educated. There is very little incentive to admit that the hoi polloi might know what they are talking about or have a good point. It is not unusual for people to be called 'anti-science' for taking positions that are extremely rational and not aligned with what the scientists want to happen. And this even though scientists are pulled from a class with its own political motivations.

Stereotypes about serfs were very accurate during feudalism.

Oppression has real effects, the core question is the direction of causality. Which way you think it points is precisely how much of an actual, factual nazi you are.

Stereotypes are by and large accurate, and unless I misread, "stereotype" in this article in not a technical term. The article uses it to mean what the layman means by it.

What do we do with this finding?

They go into detail on that in the "Knowns, Unknowns, and Emerging Controversies" section. Further questions that knowledge can help us answer include:

* How do stereotypes form?

* Why are some stereotypes not accurate? They mention nationality stereotypes in particular as being quite inaccurate.

* What other factors correlate with increased or decreased stereotype accuracy?

There's some value, of course, in just learning something's true even if it's not immediately useful knowledge.

I'd add to that: How do we keep stereotypes from reinforcing themselves?

> They mention nationality stereotypes in particular as being quite inaccurate.

National stereotypes are actually quite interesting because they come in two flavors:

"Swiss are punctual" is a regular stereotype that by and large applies to the average Swiss.

"Germans wear lederhosen" is a stereotype that is wrong for the vast majority of Germans. It's still true/useful in the sense that if you see someone wearing lederhosen they are most likely German. That makes is a good visual shorthand for "German", which can feed the impression that it's common.

On "Germans wear lederhosen", it is true that of all people who wear lederhosen, it's going to be mostly Germans. "Lederhosen are worn by Germans"

This is an argument over the definition of a word that is written to look like a carefully researched paper in sociology, psychology, and statistics.

The authors complain that the word "stereotypes" is used to refer to assumptions about groups that are incorrect (the assumptions are incorrect). They then show that by using an alternate and broader definition of the word stereotypes (to mean assumptions about groups period, rather than flawed assumptions about groups), they find that stereotypes as they define the term are frequently accurate.

This analysis is as close to a tautology as you can get while still filling that many pages with prose.

This isn't science or sociology or advanced statistics. It's a debate over the definition of a term that is used sloppily in casual language, and a debate that is buried deep in a long paper to make it look like they are doing something more important than arguing about the definition of the word stereotype.

There is a silly section on the definition of the word, but it's worth pressing past that to read the rest of the document. The core content is really interesting!

(Overall the whole thing is badly written and presented, imo. Many numbers in prose as opposed to tables; overlong sentences etc etc.)

Ahh yes, historicity masquerading as evidence of inherent attributes strikes again.

Nowhere does the article claim stereotypes are inherent. A common tactic - when confronted with evidence you don't like, switch to arguing that it doesn't support a stronger claim you made up, without acknowledging that it does support the original claim.

The original claim, stereotypes are accurate is effectively tautological given broad based, systemic reinforcement of said stereotypes.

I was, however, responding to claims upthread that were, in fact, claiming inherent attributes were responsible for accurate stereotypes.

Why are you more concerned with my statement that you believe misconstrues the original research rather than the obvious racists above who did make a claim not supported in the paper. Why?

> I was, however, responding to claims upthread that were, in fact, claiming inherent attributes were responsible for accurate stereotypes.

No, you responded to the article. If you wanted to respond to some specific comment you should have replied to it directly.

> Why are you more concerned with my statement that you believe misconstrues the original research rather than the obvious racists above who did make a claim not supported in the paper. Why?

Spare me. Even the obvious racist (whose comment is dead and already has many replies rebutting it) didn't make the claim you attribute to them. So you rebutted a claim they didn't make, in a separate comment thread, with no hint in your own comment that you were responding to anything other than the article. Then you should not be surprised you were misunderstood.

Stereotypes of serfs were largely accurate in feudal societies.

Unfortunately stereotyping has become "unacceptable" socially, and it prevents addressing issues and fixing them since talking about the issues is "morally wrong".


It appears to be a chapter from a book titled, " Handbook of Prejudice, Stereotyping, and Discrimination". I haven't read the whole chapter, but it looks like this is what is being discussed:

"It took social psychology nearly a century to recognize that not only had it been declaring stereotypes to be inaccurate on the basis of little data, but once the data started to come in, to accept that this data often (though not always) demonstrated moderate to high stereotype accuracy. This resistance to the data has constituted a significant impediment to understanding the existence, causes, and consequences of both stereotype accuracy and inaccuracy."

The authors review various kinds of stereotype research in social psychology, and say that in many cases the researched stereotypes are accurate even though researchers presume they aren't.


> very few whites are as intelligent as blacks like Neil deGrasse Tyson or Barack Obama

There is a stereotype of self-righteous closeted eugenics believers that I’m reminded of, when I read statements like this. But you go just short of pushing an agenda, so fine. Putting that aside, the thing that I find frustrating about people having this debate is that they never seem to ask “why would people with darker skin tones do worse at standard cognition tests?” Well, IMHO, a more plausible answer is that “populations that are treated and educated better, who encounter less barriers to privilege, enjoy relative ease of social mobility and physical safety, are more likely to spend time developing their abstract reasoning abilities”.

As a black person I'm going to swallow my disgust at the fact I had to read that this drivel evening and point their own study is clearly stating the opposite point they made, that IQ tests are not measuring some sort of inherent intelligence difference in races:

> Second, even IQ scores clearly respond to changes in the environment. IQ scores, for example, have risen dramatically throughout the world since the 1930s.


Pseudo-intellectual "redpilled" types like to pretend that racial conflict ended the day slavery ended, but since out redlining wasn't even fully illegal until the late 70s... it makes sense that a test highly affected by environmental factors would reflect systemic and persistent policies that were designed to destroy the environment of minorities.

As a black person I'm going to swallow my disgust at the fact I had to read that this drivel evening and point their own study is clearly stating the opposite point they made, that IQ tests are not measuring some sort of inherent intelligence difference in races

You attribute a position to me that I did not make, and do not hold. You then criticize a straw man of your own invention.

My position is that IQ scores reflect actual intelligence. However the environment that you grow up in is one of the biggest drivers of how intelligent you will become. Therefore population differences in intelligence are mostly attributable to differences in environment, not genetics.

That said, the position I hold that you don't like is that there are population differences and those population differences have a significant impact on real lives. I don't like that either. However that is what the data says, and I'm strongly in favor of accepting unpleasant realities rather than sweeping them under the carpet.

You will note that the article that I linked to supports all of my actual positions.

Let's pretend for a second, that you just don't know how to write a simple thought in a coherent way, accidentally tried to say "the stereotype blacks are stupid is true", and aren't instead just backpedaling from your original shitty comment after being called out...

You say it's a difference in environment, so why use that difference to defend a statement like "blacks are stupid"?!

Why try to make a try and enforce a racist statement like "blacks are stupid", that's actually _less valid_ and _much more easily disproven_ than the hypothesis you claim you're _actually_ backing, that environmental differences lead to differences in intelligence?

After all, while it's trivial to find studies questioning the reliability and validity (not just statistical validity) of IQ tests... it's _more_ trivial to find studies showing the effect of different cultures on IQ tests ability to even measure intelligence: https://www.semanticscholar.org/paper/Culture-Fair-Cognitive...

_The study you link is making connections to race with specific goals in mind that have nothing to do with trying to defend a stereotype as heinous as "blacks are stupid". In fact their point is literally the opposite, that what appears to be a simple gap in races, is actually a reflection of environmental differences._

It's clear to me that your own underlying bias caused you to try and draw a conclusion that literally does not exist, and you internally try to write off every single attempt to call out your own stupidity as pearl clutching.

It's literally the opposite of defending a statement like "blacks are stupid"

No testing gap being closed is continent on people "not sweeping under the rug the fact that blacks are stupid!"

Now, let me be a little personal here, you are the worst kind of person and part of a growing plague.

This new wave of people overestimating their intelligence and understanding, and undervaluing basic human decency because "it flies in the idea of logic!" and "you're only mad I said something heinous because of <insert overused logical fallacy>!!!!".

There are a lot of truly awful things that you can justify if your only criteria is "I can misconstrue a study to make a shitty defense of it!"

Do your friends at "Lean St" know that you spend your time defending statements like "blacks are stupid"? Maybe you should tell them?

I've read your rant multiple times. You haven't exactly demonstrated good reading comprehension. Nor do I believe your personal attacks are justified.

If you wish to argue that blacks do not score over a standard deviation lower on IQ tests, you're arguing against a mountain of evidence. If you wish to argue that this difference doesn't have pretty serious real-world consequences, you're arguing against another mountain of evidence.

You can choose to argue that IQ is culturally biased. I have to admit that culture affects people's IQ. But in turn IQ measures mental abilities that will help you do well in a technologically based society. So if particular cultural ideas help kids grow up with better IQs, those ideas are worth adopting. If specific family interventions have a positive impact, they are worth considering as well. And if you refuse to do that for kids, you bear a share of the responsibility for their future failures.

Seriously. IQ is a better predictor of your future success than your socioeconomic background. It is a better predictor than the color of your skin. The good side of that is that people are being judged more on their abilities than their background and race. That's good. But the bad side is that every year in which we see a large gap in SAT scores between blacks and whites, we are looking a future economic gap that will last decades.

As a society, we refuse to even think about that. Which means that we choose to continue the problem.

Read it again then, your entire comment is already answered to above.

You can keep trying to switch direction like a weasel, but I could cut and paste the comment you replied to and it'd still be accurate because you didn't answer a single point brought up.

Let's narrow the foxhole though, see if you can answer yes or no to a few simple questions:

> The answer to the implicit question is simple and obvious. The stereotypes that proved accurate were often ones which people, for political reasons, wished to be inaccurate.

> A simple example will demonstrate this. There is a stereotype that blacks are stupid.

Your comment is attempting to support the stereotype "blacks are stupid" as "accurate but inconvenient". Yes?

You must realize it's possible to talk about differences in education outcomes for different races without deciding to frame it in defense of the stereotype blacks are stupid, since you link to an article that does so. Yes?

Yet you decided to use it support the stereotype "blacks are stupid". Yes?

"Blacks are stupid" is a racist stereotype. Yes?

You are supporting racist stereotypes on HN. Yes?

You should should keep trying to double down on it? No.

Yes, I did not push an agenda. I also linked to an article that was all about how we can reduce the gap. But the very first step is to admit that said gap actually exists.

But my point is that people's current experience is going to be based on the reality that the tests reveal. Which is that the average black that they meet will, statistically, be noticeably less intelligent than the average white. And experience is going to lead to prejudices. And the fact that these prejudices are shocking and horrifying to people is obvious from the comments.

I take it that you find this frustrating. But in this political climate, I find it understandable that people are “triggered“ when you put emphasis on the current state of things, even if it is a fact, than how we ended up here. Underneath those statistics is an injustice at a large scale that span generations, and many people don’t want to build a society on top of injustice, without addressing it first. That’s not a matter of statistics, it’s a matter of values. So when you put emphasis on statistics, people will suspect that you don’t share those values.

Yes. I get that.

And they are about half wrong when it comes to me. I actually have similar values. But I have very different beliefs about how to achieve those values. And as a result, I disagree with most of the agendas being pushed to try to address the problems left by the awful history.

Yes, I want equality. But I want it to be obtained with a hand up, not a handout. We cannot successfully mandate equal outcomes for populations with unequal abilities. We've done that for 60 years and it has not worked.

In California today, black kids score a standard deviation lower than whites, and graduate high school with reading skills matching the average white 8th grader. We can't even talk about that fact. But IQ is a better predictor of your future life trajectory than the color of your skin. And those test scores guarantee that decades from now, those white kids will be enjoying better lives than those black kids. Not because of something that someone in the past did, but because opportunity is available to all, and the benefits accrue to those who are capable of more.

This isn't actually a hopeless fact of nature. Thanks to the Flynn effect, the black-white IQ gap is roughly the same magnitude as the IQ gap between white kids today and their Boomer grandparents. If black parents could learn what white parents have learned in the last 50 years, and receive support to help actually raise their kids well, I believe that in one generation we could erase the IQ gap, and most of the wealth disparity afterwards.

But nobody is chanting, "We need parenting classes!" And suggesting anything like that will get you accused of victim blaming.

So we're going to double down on the same basic ideas that have failed to solve the problem for 60 years now. Is there any reason to believe that it will start working at any point in the next 60 years?

I think your "values" are leading you to believe blatant racist and victim-blaming tropes like "black people have disparate outcomes because... they're worse parents". How about we instead start by closing the 23 billion school funding gap. Or would you consider that to be a "handout". https://www.npr.org/2019/02/26/696794821/why-white-school-di...

I see. It is racist to say things like, "We should encourage black parents to not beat their kids so much." And yet as articles like https://www.apa.org/pi/families/resources/newsletter/2017/04... point out, that happens to be a very real problem. And so we refuse to even discuss the very interventions that are most likely to have the biggest impact.

On the school funding gap, I'm all for fixing it. That's one of the root causes behind why black boys are under half as likely to meet state reading standards in California as white ones. It isn't the only one - black kids are already substantially behind before they enter school - but it sure needs fixing.

But along with that, let's not lie to ourselves about the consequences of the current situation. A black high school graduate is half as likely to have actually learned what they were supposed to as a white. A business should be able to hire high school students but require a literacy test. The fact that blacks are half as likely to pass said test should be viewed as a judgment on our school system, and not blacks.

But in our legal system, businesses who try that are likely to be sued and are likely to lose the lawsuit. (Many have.) Because insisting that your employees can read and write is considered racial discrimination.

So businesses instead require a college degree and don't have a literacy test. The result is that businesses get literate employments, they hire fewer blacks than they otherwise would have, and now there is more demand for college degrees. Which helps cause tuition to go up, and saddles young people with debt. Which means that everyone loses.

In short I support creating results. Not mandating them. What it takes to create results sounds shocking because those ideas have been off the table for discussion for decades. And likewise discussing the failures of the current system is also off the table.

The two things you've suggested as solutions are "parenting classes" and "encourage black parents to not beat their kids so much".

You're doing the same thing here as you did above with your original post.

No anti-racist would look at the results of IQ tests and frame their conclusion as "don'tcha see, black people are stupid".

No anti-racist would look at the article you posted which acknowledges that "the use of corporal punishment in black communities today is a byproduct of centuries of slavery, the racial terrorism of the Jim Crow era, and exposure to racism that continues to chip away at the vitality of black life." and come to the conclusion that "black parents just need to learn what white parents have learned".

Everyone can hear the dog-whistles loud and clear.

Because (1) the test can in fact be racist in the sense that it's optimised/over fitted toward the white (majority) population, but most importantly (2) the correlation between IQ and race is an entirely false one! Or rather, the correlation is indirect. In fact what is truly happening is that higher test scores are caused by such things as: family wealth, school, parental education, home life, etc, which black people statistically lack more than white people. Hence the correlation between IQ and race, a simplistic analysis that obscures the actual mechanisms at hand.

I always found it funny that Canadians score better (before normalization) on the WAIS IQ test than Americans even though the WAIS is an American test with some American history knowledge questions and American notation/units in the questions. Canada doesn't actually have their own IQ test afaik, they just use the WAIS and renormalize.

I'll never understand why US politicians don't strive to close this gap or many others when the two countries are otherwise very similar.

Taking a single statistic on something as complex and multivariate as "intelligence" is already discarding a lot of information. When you throw in social factors (Who is writing the tests? What are the challenges faced by the various populations taking the test?), it becomes utterly absurd and unscientific to conclude that your statistic is free of some systemic source of error.

If IQ tests were presented in, say, a physics context as a serious experiment, the presenter would be laughed out of the room. Taking them seriously and uncritically indicates nothing less than a complete lack of understanding of how the scientific method is to be carried out and how its results are to be interpreted.

Citation for for IQ test results == intelligence?

While not perfect, they are the best metric we currently have.

Wikipedia says "In psychology, human intelligence is commonly assessed by IQ scores that are determined by IQ tests. However, there are critics of IQ tests who, while they do not dispute the stability of IQ test scores, or the fact that they predict certain forms of achievement rather effectively, do on the other hand argue that to base a concept of intelligence on IQ test scores alone is to ignore many important aspects of mental ability.

On the other hand, Linda S. Gottfredson (2006) has argued that the results of thousands of studies support the importance of IQ for school and job performance. She says that IQ also predicts or correlates with numerous other life outcomes. In contrast, empirical support for non-g intelligences is lacking or very poor."


> IQ also predicts or correlates with numerous other life outcomes.

If we hire people based on IQ scores/things that correlate with IQ scores, we should not be surprised if high IQ leads to better outcomes.

correlation != causation

We have evidence of causation here. Research during the 1960s shows that IQ tests do a better job of predicting future job performance than other measures, such as interviews. As a result, many companies began to use IQ tests for hiring.

This ended with https://en.wikipedia.org/wiki/Griggs_v._Duke_Power_Co.. The fact that IQ tests resulted in hiring whites at a higher rate than blacks was deemed a violation of the Civil Rights Act and so the practice was banned. The correlation remains true, but it is illegal to hire based on IQ.

Do IQ tests control for home life stability/quality, parental and generational wealth/education, school quality, other forms of hardship?

No, IQ tests are sheets of paper full of questions like "complete this two-dimensional sequence of shapes", "hand = finger, foot = ?", etc. that get scored based on your answers.

Studies on IQ tests however do control for the factors you named, at least if they are good studies.

More importantly, the IQ test was created specifically as a way to validate racism.

Stanford psychologist Lewis Terman’s IQ test, the standard upon which our current tests are based, was meant to show “enormously significant racial differences in general intelligence, differences which cannot be wiped out by any scheme of mental culture.” The Measurement of Intelligence 1016


Are you saying that he is factually wrong or that it is merely socially unacceptable?

He is wrong, as pointed out in other comments. I'd rather him not delete it though.

I'm sure that I said things you did not like. But what, specifically, did I say that was wrong? Can you identify a falsehood? Please be specific.

And before telling me that it is false, please read the link that I gave. Digest the information in that link. And be sure that you're criticizing me for something that there isn't data in support of.

Are you asking because you think the parent point is right? (despite dozens of replies to the contrary)

Because your wording strongly implies you understand it's additionally repulsive...

It's obviously repulsive. It starts with how quoting evidence for some stereotypes is career suicide, then continues arguing for just such a stereotype "just to give an example".

However note how the comment I replied to was the first response, and at the time I replied the only response. I like to imagine that hacker culture is still about judging ideas on their merit. Asking someone to delete their statement instead of refuting anything they say is wrong by that standard.

When someone uses overtly racist language and is making a blatantly ignorant statement, then I don’t think I need to give details other than to try to prevent others from reading it.

> I like to imagine that hacker culture is still about judging ideas on their merit.

If by hacker culture you mean Stallmanist culture, sure. Stallman is the guy who thinks necrophilia has merit because of a wrong interpretation of the Constitution, and therefor, being anti-necrophilia is narrow minded.

He and his followers are true paragons of merit-based discourse.

Now... for the rest of us, some ideas are blatantly disdainful, you can skip explaining why they have no merit and go straight to telling someone to delete them.

I mean seriously, if your comment sounds like some sort of eugenics manifesto, just delete it, why do you deserve an explanation of why that's bad?


This whole comment chain reminds me of the new wave of anti-anti-racism where "racism is ok if you support it with well misconstrued factoids"...

The problem is, there's almost no overlap between ideas that are so blatantly racist they don't deserve a second thought, and ideas that are worth a second thought upon examination.

See: above comment's useless point, which was wrong anyways, or Stallman

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact